File size: 947 Bytes
cfbc06c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32

- datapaths: ["chr20_21_22.fa"]
  output_dir: "./preprocessed_data"
  output_prefix: chr20_21_22_uint8_distinct
  train_split: 0.9
  valid_split: 0.05
  test_split: 0.05
  overwrite: True
  embed_reverse_complement: true
  random_reverse_complement: 0.0
  random_lineage_dropout: 0.0
  include_sequence_id: false
  transcribe: "back_transcribe"
  force_uppercase: false
  indexed_dataset_dtype: "uint8"
  tokenizer_type: "Byte-Level"
  vocab_file: null
  vocab_size: null
  merges_file: null
  pretrained_tokenizer_model: null
  special_tokens: null
  fast_hf_tokenizer: true
  append_eod: true
  enforce_sample_length: null
  ftfy: false
  workers: 1
  preproc_concurrency: 100000
  chunksize: 25
  drop_empty_sequences: true
  nnn_filter: false  # If you split your fasta on NNN (in human these are contigs), then you should set this to true.
  seed: 12342  # Not relevant because we are not using random reverse complement or lineage dropout.